ByPassAuth is essentially an “AI stealth cookie session tester.” It’s built to take a Netscape-format cookie file, try those cookies on a list of websites in a stealth Chrome session, and then use GPT-4o vision to decide whether the page looks like you’re logged in (YES) or not logged in (NO).

Here’s what it does, step by step, in plain English:

What it’s for

• You give it a cookies .txt file (Netscape format) and a comma-separated list of sites.

• It opens each site in an “undetected” Chrome browser (tries to look less like automation).

• It injects any cookies from your file that match the site’s current domain.

• It refreshes the page and takes a screenshot.

• It sends that screenshot to OpenAI (GPT-4o) and asks: “Is the user logged in?”

• It sorts the result into valid (looks logged in) or invalid (looks logged out), saving screenshots as proof.

The main pieces in your script

1) Stealth browser automation

• Uses undetected_chromedriver to launch Chrome with automation signals reduced (e.g., disabling AutomationControlled).

• Requires a local chromedriver.exe in the same folder, and it must match your installed Chrome version.

2) Cookie file parsing (Netscape format)

parse_netscape_cookies() reads the cookie file line-by-line and converts it into dictionaries like:

• domain, path, secure flag, expiry, name, value

3) Cookie injection (domain-aware)

add_cookies_to_driver():

• Looks at the browser’s actual current URL after redirects.

• Only injects cookies whose cookie-domain matches the site domain (e.g., .google.com should apply on mail.google.com).

• This is specifically meant to handle situations like gmail.com redirecting to accounts.google.com.

4) Screenshot → AI decision (logged in or not)

analyze_login_status():

• Takes the screenshot, base64-encodes it, sends it to the OpenAI Chat Completions API.

• The prompt forces a strict YES/NO based on visible UI clues like:

  • profile avatar

  • sign out button

  • dashboard pages

  • vs. login/sign-in pages

5) Output organization

It creates a timestamped folder like:

• Scan_Results_YYYY-MM-DD_HH-MM-SS/

  • valid/

  • invalid/

For each site:

• Saves the screenshot in valid or invalid.

• If valid, it also copies the original cookie file into the valid folder renamed per site.